McKinsey & Company recently sat down with Stuart Russel—renowned computer science professor at the University of California, Berkeley and “one of the pioneering thinkers” in artificial intelligence (AI)—to discuss AI technology and its future. In a May 2020 McKinsey podcast, Russel discusses the meaning of AI, how to ensure that it benefits humanity, and the concept of “super intelligent” machines, providing a fresh perspective on the impact of this groundbreaking technology. Below we look at some highlights from the discussion. Enjoy!

The Meaning of Artificial Intelligence

According to the podcast, there’s lots of hype, but also lots of misunderstanding around AI. Here’s how Russel defines it: “AI is about building machines that do the right thing, that act in ways that can be expected to achieve their objectives. This covers learning systems, robotic systems, the game-playing systems, the natural-language systems—they can all be understood in this framework.” According to Russel, most AI in use today is logic-based, meaning they use business rules or business intelligence systems, and are aimed at predicting things that may happen next. Examples include determining which product a customer is most likely to purchase next, predicting when machinery and equipment need maintenance, and automating subscription renewals.

Smaller Goals Often Lead to Bigger Breakthroughs

In the podcast, Russel makes the case that major advancements in AI typically result from less ambitious, more narrow projects aimed at solving a simple problem. For example, back in the 1990s, AT&T Labs worked on an AI solution to recognize handwritten digits that could be used by the U.S. Postal Service and banks to recognize handwritten checks. According to Russel, this relatively “mundane” goal led to the development of convolutional neural networks or CNNs, a major AI breakthrough that enables users to mine intelligence from unstructured data such as images. Common examples of CNNs in action today include scanning health-related images to better diagnose diseases or detecting product defects using images and visual analytics.

A Future of Super Intelligence?

Russel also discusses “super-intelligent” machines and how they may play a role in our future world. By definition, super-intelligent machines bring more general and far-reaching impact from AI initiatives than more narrow, business-focused use cases. This includes AI systems that can mimic the way the human brain learns, reaching the same level or greater levels of intelligence than people. When asked if “abandoning” this level of AI technology was the best way to prevent it from taking over the world, Russel stressed the need for humans to maintain control over super-intelligent machines. To do this, the focus of AI development needs to shift from “intelligent” systems to “beneficial” ones—meaning systems should be designed to help humans, not harm them.

For the complete discussion with Stuart Russell, including how to shift our focus to ensure beneficial AI, read the McKinsey & Company article and listen to the podcast.